You're tuned in to Tech Beat. Here's what's moving in the world of technology today.
In a result few saw coming, OpenAI's GPT-five point five has edged out Anthropic's Claude Fable five on a demanding new benchmark called Agents' Last Exam. Built by researchers at UC Berkeley alongside more than three hundred domain experts, the test is designed to measure whether AI can actually complete real, economically meaningful professional tasks from start to finish — not just answer questions. It's a meaningful shift in how we measure machine capability, and the upset has the AI community paying close attention.
Meanwhile, New York has quietly become a pioneer in AI transparency law. Advertisers in the state must now clearly label any AI-generated synthetic performers appearing in their ads. The regulation puts a legal obligation on brands to disclose when the human face or voice you're seeing isn't actually human — a modest but significant step in a landscape where that distinction is growing harder to make.
And over at Xbox, CEO Satya Nadella's gaming division is facing a moment of reckoning. In a public letter to staff, the Xbox chief acknowledged that current profit margins simply cannot continue, signaling that structural changes are likely ahead for a business still searching for its footing in a fiercely competitive gaming market.
That's your Thursday briefing. Keep surfing. Tech Beat out.
